Switching Regulator

[Noggin]

In my senior design group, I gave my group mate a schematic to build our power supply for our project. It consisted of a 7805 and a PNP (or NPN, I forget which) pass transistor. He replaced the BJT with an opamp and an N-FET. I was quite annoyed with it at first because while doing all of that, he also made the rest of the circuit much more complex.

So did he basically make a more efficient 5v source by doing this? We'll need about 2 amps, and with what I had quick math shows that the BJT would have to dissipate about 18 watts (its going in a car, changing 13.8v to 5v, about 2 amps current). But if his is truly a switching supply, it would reduce the power consumption right?

The opamp - is connected 5v from a 7805
The output of the opamp goes to an N-FET
The N-FET source is connected to 12v to 13.8v
The N-FET drain is connected to the opamp + and is used for a 5 volt 2 amp (6 amp capable, though fused much lower than that) supply

I used an oscilloscope and saw as much as a 800 mV ripple on it! but I'm not certain I had it set right, it was on AC coupling. Not one darn TA I've talked to knows what the hell to do with an oscilloscope, nor can they point me to a user's manual. The only thing I know to do is hit the darn "autoset" button.

Might need to use a bigger cap on that 5v line.... I got a .5 Farad cap, that should be plenty, but I don't think I can solder it to my circuit board...

 

[Roff]

Deja vu, Noggin. Like I said before, why don't you post a schematic? Your description is way too sketchy to understand.

 

[laroche73]

regulator

Could you post a schematic? The 7805 is a linear regulator and changing the pass element from a BJT to an op-amp/FET combination won't make it any more efficient. The pass element still needs to dissipate most of the 18 watts.

Why do you think this is a switching power supply? Are there more circuit details I'm missing?

 

[Noggin]

Deja vu, Noggin. Like I said before, why don't you post a schematic? Your description is way too sketchy to understand.

Oh man I'm sorry.... I did a quick look for that but I missed it. Didn't even think to run a search, I dimissed it as too long and just a ranting post and didn't put my consideration into it.

I'll see if I can come up with a schematic...

As for why I think it may be a switching supply, its based only on the fact that the NFET doesn't seem to ever get into a steady state. It just keeps wobbling back and forth... can't think of a better word.

That opamp is a 741. Not sure about the PFET, wouldn't think it matters though as the op-amp would adjust accordingly, rated for 8 amps. I believe we have a 470 micro farad cap on the drain of the PFET.

I need to install a schematic program, maybe I'll see if I still have PSPICE somewhere...

 

[crust]

Perhaps I am not clued in, but I have no idea how this circuit works even after looking at the schematic. It seems that you have somehow configured the opamp as a follower. This is in no way a switching power supply. The 7805 is a linear regulator. if you are drawing 1A, presumably the 7805 will dissipated 1A at 13.8V-5V or about 8W. Perhaps you can provide a clearer idea of what your end goal is: For instance, are you attempting to construct a 13.8-5V switching power supply? Or are you trying a high power 5V linear power supply (the 7805 is rated at 1A or so, do you want a 5V/5A?)

 

[spuffock]

You have built an oscillator. OOPS :shock:
Go back to the first idea, save money and gnashing teeth.

 

[Noggin]

You have built an oscillator. OOPS :shock:
Go back to the first idea, save money and gnashing teeth.

haha no wonder there seems to be an 800 mV swing on that 5v line. I'll strongly suggest that we put that BJT circuit on a breadboard if we determine that the 800 mV ripple does exist and we cant greatly reduce it. However, the cellular modem is able to operate off of this power supply.... I'm not completely certain that the 800 mV ripple is there.

As for if I'm trying to build a switching supply, no... just want something that works. The 7805 should dissipate VERY little heat. Its ouput ONLY goes to an opamp, which of course has very high input impedance. The ouput of the NFET goes to the other input of the opamp, and the output of the opamp drives the NFET until there is 0 difference between the 7805 and the FET drain.

I actually did put it into PSPICE last night, those are ideal parts in PSPICE, and it showed 0 ripple. I don't know how to change those things so that they're not ideal.

Here is a better shcematic of it though... if anyone can help me add some ripple on the 13.8v DC source I'd appreciate it. I tried several AC sources, but couldn't figure out how to do an appropriate AC sweep on it. Instead of a sweep, I'd rather just select the frequency and then run it at the frequency for some set amount of time though.

On the graph, the teal line (wait, I'm a guy... pretend I said blue! I don't know what teal is I swear!) is the 5v rail, and the mauve (I MEAN PURPLE! :shock is the output of the op amp.

Also, the higher the gain of the opamp the closer it gets to 5v (well, 4.96 since that is what our 7805 outputs).

Edit: The resistors are meant to just be a load on the circuit. R2 is the load from 0s to .1s, R1 || R2 is the load from .1 to .2, and R2 is the load from .2 to .3 seconds. This is just the approximate load for when the modem is or is not transmitting.

 

[Roff]

You do in fact have a negative feedback loop, but it's unstable due to multiple poles at low frequency. If you really want to use a linear regulator, look at LM350K mounted on a heat sink. I personally would use a switching regulator unless the 2 amp load is low duty cycle (<20%) and of short duration (a few milliseconds).

 

[Noggin]

Well our design specification calls for 6 amps, though we should never go above 2.5 amps. However, the 2 amps will be drawn for as long as 30 seconds, not sure of the duty cycle though. But because we'll be drawing that much current for that long, we wanted the supply to be overbuilt considerably.

Our old modem required 4 amps @ 12v.

Even with that LM350 we'd need another pass element. I think the best bet is to go back to the BJT, whether Matt understands it or not doesn't matter.

I don't quite understand the terminology of "negative feedback loop" and "multiple poles at low frequency." I lean strongly towards the world of digital, and did poorly in my physical electronics course (well I consider a 'C' to be poor) and don't have much hands on experience yet in this field.

Is there a decent switching supply you can recommend? I'll search digikey for now...

 

[Noggin]

LM2678 - https://www.electro-tech-online.com/custompdfs/2003/10/LM2678.pdf

I don't understand something, why have all the parallel caps?

 

[Sebi]

These capacitors are low ESR -type.Todays you can also purchase one 47uF and 360uF (or 470uF).

 

[Noggin]

Ok, so a single big one on the input and output should work then... Its supposed to be DC so I imagine its just to reduce ripple, so the only drawback to one being too big is cost....

Just heard something funny, gotta add it to my sig...

I wish there was an off topic section here. I live in off topic forums, and its hard for me to stay on topic sometimes (like right now.)

 

[Noggin]

Ok, been doing some reading from google...

Negative feedback loop, output is used as an input to adjust itself. Pretty simple concept...

Well I just got back an email from him, he won't agree to using it. He said that he didn't feel the inductor is ideal.... like anything we have is ideal.

 

[seeker]

Hi Noggin,
Seems to me you want a simple 5 volt regulator supplied from a car's 12 volt system.
A plain LM7805 driving a pass transistor (bipolar or FET) with suitable heatsinking should do it. You should not have any trouble with ripple in a car.
You can also jack up the output of the 7805 a little by adding plain old 1 amp rectifier diodes in series with the ground lead of the 7805. How precise does the regulation have to be?

 

[Roff]

Well I just got back an email from him, he won't agree to using it. He said that he didn't feel the inductor is ideal.... like anything we have is ideal.

Yeah, as if wasting 18 watts in a series pass regulator is ideal. Remind him that an automobile is not the ideal environment for getting rid of 18 watts of heat. A switching regulator can be over 90% efficient. Deliver 10 watts to the load and waste less than a watt.

 

[Noggin]

How precise does the regulation have to be?

Acceptable ripple voltage for the GPS unit on Vdd is 150 mV, the microcontroller (Atmel ATmega128) and modem (Multitech GSM SocketModem) are unspecified.

We put it on an oscilloscope today, he knows how much I hate it.

With no load, there is about 30mV ripple directly at the output of the NFET with a 470 uF cap (using a breadboard).

With the microcontroller on the circuit board (NOT the breadboard) there is a 30mV ripple at the output of the NFET with a 100 uF cap, and 800 mV ripple at the GPS unit. If a 470 uF cap is added at the GPS unit, the ripple drops to 120 mV which is acceptable.

We brought an o-scope outside to the car to see if engine noise was going to make a huge difference, but the outlet breaker was apparantly thrown, couldn't get power for the o-scope. I have a 140 watt power inverter that should run the o-scope though. We'll try to check that tomorrow.

If we strong arm it with caps, this may work. We have until the first or second week of December to finish this project. Although I hate his power supply design, I'd hate to not have a working unit because we didn't have time to put this one together.

Anyway, I just figured up that the OPAMP/NFET combo supply is going to cost about $12 once everything is beat down with caps, and the LM2678 supply is going to cost about $14 with the suggested caps and a sturdy inductor.

I don't even know if that would fix the problem though, it seems that even if we don't see noise at the NFET, we still have a huge ripple at the GPS unit. It looks like the microcontroller is making it.... the traces are already pretty darn wide though.

 

[UberGeek]

Noggin,

I'm casting my vote for the LM2678 solution...especially if you're powering a micro. That regulator is specifically designed for the automotive environment which can be quite ugly...huge voltage spikes from load dumps and all manner of nastiness. There are other similar regulators from National that you can find with a parametric search. Oh...and read the whole datasheet for the LM2678, it's full of good info. They have an excellent app note about automotive LDO's as well. LDO regulators have very specific capapcitor ESR requirements, hence the multiple input caps...all explained in the datasheet.

 

[Noggin]

We had a meeting with our advisor today, I couldn't make it. Matt went, and he just called me on the phone. Matt said that Dr. Reese suggested a few things, but that he also seems to be leaning towards us to use the LM2678. Tomorrow, we're going to go into the high voltage lab and hook this thing up to a Humvee and see how ugly it looks.

Also, I'm going to call Digikey and ask for samples for the regulator, inductor, and caps.

I'll take a thorough look at the regulator pdf again. Thanks

 

[Roff]

Mmmmm... Does Digikey really give out samples? I think you'll have better luck getting them from the manufacturer.